Re: different explain-lite comment

Bob wrote:

> so why have numeric USE values?  And why have attribute sets? 
> I have been debating this for myself - once you have EXPLAIN
> and the server tells the client what attributes go with 
> the user selected search (e.g. "title") and these are sent back - well then
> why bother? Just tell the client the allowed searchname/field names that
> are searchable. And have what the user selected sent back! But I never bring
> this issue up - violates all sorts of things - [snupped]

Its probably not worth spending time on as its not really likely
to change, but we do find having one layer of field names for users
(our CCLInfo) and another layer at the system level (Bib-1) useful at
times. What Bob is suggesting is more like a SQL model (hence most people
feel more comfortable). But by having a numeric identification scheme
you can apply multiple overlapping attribute sets to provide different
views of the same data (Bib-1, STAS, GILS, ...). These different sets
may have concepts of "title", but they are different. The CCLInfo 
definition can map "names" onto "combinations of attribute concepts"
(hmmm: maybe a CCL field name in CCLInfo should allow multiple attribute
lists to be specified in an OR group). I guess different CCLInfo
structures could also be returned for different types of users for the
same underlying database giving different end users different views
of the database (eg: novice gets author, title, subject while advanced
gets lots of fields).

But I struggle to make arguments I feel are convincing even to myself.
I think numeric USE attributes are so deep seated that its not even
worth thinking about changing it unless Z39.50V4 came along with a
radical revamp of the protocol. So I probably should not have sent
this email! :-)

ajk

Received on Thursday, 23 November 2000 18:42:57 UTC